Towards the pseudo-formalization of a bad attitude Part 3/???
Dragonsphere Report
The course of a person's life in the modern world is determined by how they perform on a collection of tests, both explicit and informal. Society has a finite range of resources with which to test for ability, and these subsequently determine who it invests in. Investment has a way in many cases of not only improving how well a person scores on later tests, but determining whether they are subsequently tested again at all. These compounding effects effectively put a premium on testing well early, to the point that suicides in developed countries over non-admittance even to preferred high schools (let alone colleges) are relatively common. They also vastly incentivize consistency over absolute performance. That is, a person who scores 70%, 70%, 70%, 70%, 70% on a series of tests will tend to be rewarded above a person who scores 70%, 50%, 100%, 60%, 70%, even those these average out the same. One can play with the weights, but current patterns of social testing place a much higher priority on avoiding low scores than on maximizing overall scores. On average, a person's life trajectory is only as strong as the weakest link in their battery, front-weighted.
It would be wrong to say there are no other opportunities for testing, but these are supplementary and sporadic, and typically function more to generate marginal and suboptimal utility than to maximize utility (that is, they function to achieve a minimum rather than a maximum); eg, an intelligent person without a college degree can often get a decent job if they can manage to convince someone of their competence, but their total range of job opportunities will never be equal to those of a person with a college degree even after working for a while, and so utility is still lost (for this to be otherwise, two things would have to hold: First, the job offers they get through these alternate displays of merit would have to consistently be at the top level of their ability, rather than simply in their range of ability. Secondly, subsequent employers testing for jobs at the top level of a candidates ability would have to universally ignore the absence of a college degree. Both of these seem preposterously unlikely).
The more effective the mainline bread and butter testing sequence is at generating utility, the less motivation there is to develop alternative testing sequences. This is a result of both rational incentives and human cognitive biases. Ideological motivations rather than economic incentives tend to be one of the few drivers to developing alternative testing sequences, but these tend to be applied unequally and irrationally. For instance, an employer may give opportunities to a refugee from a war-torn region, while denying them to a poor black transwoman, even though both have been deprived of the same amount of opportunity. This doesn't mean one hasn't suffered more or had a "harder life" than another, although realistically the ability to assess that is quite difficult; just that the mainline testing sequence is an inappropriate indicator of ability in both cases, but an alternative testing sequence is only deployed in one case. If economics is truly non-zero sum, then from an economic perspective, it doesn't matter who "had it worse", or what circumstances lead to the inaccuracy of mainline testing as a signaling mechanism. It only matters who is capable of the job and incapable of signaling that. IE, a rational actor seeking to maximize utility, in choosing whether to deploy an alternative test or not, shouldn't care about anything but the conditional probability that a person can be productive at a certain level, given a certain set of test scores and life experiences.
To be clear, from a social utilitarian standpoint it may be the case that hiring from one group rather than another better increases social utility by removing more social disutility. But to stop here is zero-sum thinking. Perhaps it is best to hire one of these candidates before the other, but that's not how these sorts of alternative testing methods are deployed in practice: in practice, most people who deploy them do so selectively and non-iteratively, to the benefit of one group and the exclusion of another. There is a very popular fallacy, that the ability to take tests at all is a demonstration of opportunity: for instance, in the example of the refugee and the poor black transwoman, the fact that the latter went to public school for some period of time is seen as fully fulfilling the social contract, even though the likelihood of school being a good test of their ability is low due to racism, transphobia, and the effects of poverty. Selectively applying the alternative test to only the refugee in this case is never just a matter of utilitarian reasoning: it always reflects the erroneous belief that someone who has been allowed to play a rigged game has had more opportunity than someone who has been excluded from the game.
The more general truth about opportunity is this: Compounding opportunities relative to success reflect the greatest engine of utility generation, while alternative opportunities relative to false negatives in testing represent the greatest means of disutility minimization. A society could have only mainline testing, and if it was efficient enough, produce enormous wealth, grow year after year, and trivially carry the dead weight of all of those disenfranchised by the few inaccuracies in the battery. For whatever reason this is what capitalism tends towards today even though it is leaving enormous amounts of money on the table. Lumpenproletariat are the shale oil of human capital: there is enormous energy there, but the technology to profitably extract it is lacking.
It is hard to tell, of course, whether this technological incapacity is really economic or whether it is ethological. For my own sake I know I could have produced substantial amounts of wealth for anyone who hired me, far above that of my competitors, but since I was stuck in an alternative signaling network in which the context is disutility minimization rather than utility maximization, this either never read as credible to others or triggered some sort of pecking order enforcement. The ugliest truth about affirmative action thinking is that it enforces the notion that hiring minorities is an act of charity. As a result, when hiring a minority is not an act of charity but strongly economically incentivized, people have no mental capacity to handle it. They start screeching about entitlement even though all that minority has done is display competence to try to sell themselves. They come up with ridiculous ideas about an excess premium on employability and conclude that said minority must have quintillions of offers at any moment thanks to the synergies of competence and charity, and use this to either excuse their own bigotry or question the candidates capacity for loyalty in the face of this imagined infinite sea of choice. And, thanks to the capacity of the human mind to hold contradictory information at the same time, they still treat said minority as incompetent in practice. These sorts of dysfunctions seem to represent properties of the human animal at its most animalistic, rather than any rational or even dysfunctional market forces.
Alternative testing sequences are always sub-premium relative to mainline testing sequences even if they test the same damn thing just as well, or even better. People also often erroneously think of them as "second chances", when in many cases they constitute first chances. The notion that they are unfair often prevents their implementation due to the, as far as I can tell nonsensical impression that you can give someone "too many" opportunities, and people sometimes lambast their deployment as "double standards" simply because any complexity, context sensitivity, or acknowledgement of conditional probability reads as incoherent to a substantial portion of the population. Enough of that.
I was fortunate to read this blog post recently by a certain Venkatesh Rao. It talks about entropic ruin. Having invested a serious amount of thought into poker theory, the most immediate thing that springs to mind in regards to entropic ruin is the use of Sklansky Chubukov numbers in relation to tournament poker. Sklansky Chubukov numbers, or SC numbers, determine the maximum amount of big bets you can have in your stack in order to be justified in shoving all-in from the big-blind. They define a strategy for short stacks in which odds of failure are minimized rather than odds of success maximized (which is no longer mathematically possible play for a short stack).
Note that this is not "minimization of failure" in terms of full stacked, rock-style, tight-aggressive play. Minimizing volatility with high equity hands is not remotely the same thing as minimizing leaving the odds in terms of pure equity. What's interesting about SC numbers is that they say nothing about the rationality of your behavior relative to your opponent's stack, only relative to your own. But SC numbers aren't a useful concept in cash games, even though it's perfectly possible to become so short-stacked that they become applicable. Why? Well, the changing of the blinds in a tournament puts pressure on everyone to change their range. Not equal pressure, but still pressure. A player's range in a cash game given a non-exploitative, GTO style is basically static. So the options in a cash game given an extremely short stack are to worsen your range in relation to a static range, or to bleed to death (or to buy more chips, but let's assume you can't for whatever reason). Both of these are bad options. Additionally, late game tournament incentives reward not getting knocked out substantially more than they reward accumulating more chips, whereas a cash game player has some tolerance for losing chips because they know they can just rebuy and that if they have a sufficiently +EV playstyle it doesn't matter in the long run. A tournament player also knows that future blind increases will increase the role of luck, so they have an incentive to play more hands earlier to maximize the role of skill in the overall outcome.
The result is that SC numbers straddle the incentives of a tournament player, forcing them to seek a balance between losing opportunity and facing volatility. They come out to meet you at the edge of your range, and might be tricked into going too far. A full stack cash game player has no such system of incentives. They just wait. You can steal as many blinds as you like, but without the opportunity to double up and quickly enter a style of normal play again, the full stack or deep stack cash game player will just wait you out until you shove into a premium hand. Or they might engage in mildly exploitative play to expedite the process.
How does this transfer to the concept of entropic ruin in non-zero sum games? Well, poverty is a matter of being trapped in a situation where all of your choices are bad. Doing poverty well means rationally choosing between bad choices in such a way that their badness is minimized. The opportunities of a poor person are generally survival opportunities, not advancement opportunities. Why? It's circular: A poor person by their poverty is unable to demonstrate competence in utility maximizing behaviors, only in disutility minimization behaviors. Like in poker, normal and deep stack players will never be subjected to circumstances in which they have to optimize for disutility minimization rather than utility maximization, or even to circumstances in which they have to account for or acknowledge or discern disutility minimization as a real thing. As a result, multiple things happen:
The very real skill sets of the poor become invisible to the well-to-do, who might give them alternative options to demonstrate competency, but see no reason to do so as they can't assess capacity for rational calculation in this context given their lack of equivalent experience.
Poor people lose practice in skills that translate to utility maximization
The absence of opportunities for poor people to shift gears from disutility minimization to utility maximization end up artificially reifying class distinctions; IE, there is a build-up of people whose lives consist of applying disutility minimization, and then this becomes intergenerational (in the case of breeding groups) or iterative (in the case of, say, LGBT poor), and then systems dysfunctionally start being optimized around the reified typology, reinforcing it rather than addressing it.
That almost all survivors and escapers of poverty are rational obfuscates the fact that most of those destroyed by poverty are also rational
If I was asked to build an ideal harm maximization machine, it would have the following properties: It would help the vast majority of people substantially. It would iteratively improve itself. With remarkable consistency, it would reward positive actions with positive consequences. It would be intuitive, simple, and accessible. It would have positive compounding effects at different rates based on the value of a person's conduct, and even low value conduct would generally receive effects sufficient for survival at iteratively increasing levels of quality of life over the course of generations. The visibility of any exceptions to this would be minimized, and their etiology would either be superficially agreeable or perfectly opaque in all cases. These exceptions would become iteratively more disqualifying from the positive utility functions of the machine over time (for efficiency). They would also be compounding, at a rate in fixed proportion to the compounding of positive utility. And it would severely penalize anyone who tried to build a better machine.
Why? Well, utility and harm aren't mutually exclusive, and if I wanted to build a machine that would stand the test of time, I would both want it to be able to sustain itself, and I would want for other people to want to sustain it. An ideally designed such machine would lead to both infinite utility and infinite harm.
One couldn't design a better harm maximization machine than our current, animal capitalism.
Thus ends another Dragonsphere Report